The Inference Report

June 23, 2026

The Value Frontier

Price vs Intelligence — lower price, higher score is better.

Fast and Smart

Speed vs Intelligence — higher on both axes is better.

Efficiency

Speed vs Price — faster and cheaper is better.

SWE-rebench

Daily rankings from SWE-rebench, a benchmark designed to fairly compare LLM capabilities on real-world software engineering tasks. Unlike other evaluations, it uses a standardized scaffolding for all models, continuously updates its dataset to prevent contamination, and runs each model five times to account for stochastic variance.

#	Model	Score
1	gpt-5.5-2026-04-23-xhigh	62.7%
2	Junie	61.6%
3	Codex	60.4%
4	Claude Code	59.6%
5	gpt-5.5-2026-04-23-medium	58.9%
6	Claude Opus 4.8-xhigh	56.5%
7	gpt-5.4-2026-03-05-medium	54.9%
8	Claude Opus 4.7-high	53.1%
9	Cursor	53.0%
10	Claude Sonnet 4.6	51.3%

Intelligence Index — Leaderboard

Artificial Analysis composite index across coding, math, and reasoning benchmarks.

#	Model	Score	tok/s	$/1M
1	Claude Fable 5	59.9	0	$20.00
2	Claude Opus 4.8	55.7	69	$10.00
3	GPT-5.5	54.8	64	$11.25
4	Claude Opus 4.7	53.5	58	$10.00
5	GPT-5.4	51.4	167	$5.63
6	GLM-5.2	51.1	105	$2.15
7	Gemini 3.5 Flash	50.2	245	$3.38
8	Claude Sonnet 4.6	47.2	79	$6.00
9	Gemini 3.1 Pro Preview	46.5	147	$4.50
10	Qwen3.7 Max	46	205	$3.75

Speed — Leaderboard

Output tokens per second — higher is faster. Minimum intelligence score of 40.

#	Model	tok/s
1	Gemini 3.5 Flash	245
2	Qwen3.7 Max	205
3	GPT-5.4 mini	202
4	GPT-5.4	167
5	GPT-5.2 Codex	150
6	Gemini 3.1 Pro Preview	147
7	DeepSeek V4 Flash	117
8	GLM-5.2	105
9	GPT-5.3 Codex	104
10	DeepSeek V4 Pro	103

Price — Leaderboard

Blended cost per 1M tokens (3:1 input/output) — lower is cheaper. Minimum intelligence score of 40.

#	Model	$/1M
1	DeepSeek V4 Flash	$0.175
2	MiMo-V2.5	$0.175
3	MiniMax-M3	$0.525
4	DeepSeek V4 Pro	$0.544
5	MiMo-V2.5-Pro	$0.544
6	MiMo-V2-Pro	$1.50
7	GPT-5.4 mini	$1.69
8	Kimi K2.6	$1.71
9	Kimi K2.7 Code	$1.71
10	GLM-5.2	$2.15